This paper presents a novel measure of subtle government intervention in the news market achieved by throttling the Internet. In countries where the news media is highly regulated and censored, the free distribution of information (including audio and any visual imagery) over the Internet is often seen as a threat to the legitimacy of the ruling regime. This study compares electoral outcomes at polling station level between the Russian presidential election at the beginning of March 2012 with the parliamentary election held three months earlier in December 2011. Electoral regions in two cases are compared: regions that experienced internet censorship at the presidential election but not the parliamentary election; versus regions that maintained a good internet connection without interference for both elections. Internet censorship is identified using randomised internet probing data in accuracies down to 15-minute intervals for up to a year before the election. Using a difference in difference design, an average effect of increased vote share of 3.2 percentage point for the government candidate is found due to internet throttling. Results are robust to different specifications and electoral controls are used to account for the possibility of vote rigging.